The present invention relates to management of virtual machines (VMs), and more specifically, to a relative resource allocation among guests.
Service providers (e.g., cloud computing or virtual machine service providers) typically promote an availability of resources that exceeds a threshold value. When resources, such as storage or memory, are overcommitted (e.g., when the number or amount of resources requested exceeds actual resource availability), frames must be taken from guests in order to keep a supply of frames available for use by the system and other guests.
When taking or de-allocating frames from one or more guests, a determination may be made regarding how many frames to take from each guest. The determination may be based on an attempt to de-allocate frames from guests “fairly.” One or more variables may be considered in the determination. For example, the relative sizes of the guests, the sizes of so-called “working sets” of the guests, the activity levels of the guests, and a demand for frames in a system may be taken into consideration.